Fast User Space Mutexes (aka Futexes)

[Hiten Pandya]

Bill Huey (hui) wrote:

On Tue, Oct 21, 2003 at 04:05:04PM +0100, Hiten Pandya wrote:

Hi Gang,

Sinces the day when Matt was interviewed by the SlashNET people, a 
couple of questions arised as to the implementation of Futexes for the 
new libcr which is going to be written.

I am sending this mail as an opening to this discussion.  Hopefully, 
some of the Linux seniors scanning this lists (Rik et al) will provide 
some comments and suggestions on this topic.

IMHO, from the way Futexes work, it would be great to have an 
implementation of it for DragonFly.  For those of you wondering what 
Futexes are exactly, please refer to [1], as thats the most complete 
documented information I could find, apart from the Ottawa proceedings 
(Rusty Russell).

[1] - http://www.catfive.org/cgi-bin/man2web?program=futex&section=4


http://people.redhat.com/drepper/futex.pdf

Better documentation on how it's used and some implementation issues with
glibc. The problem with futex's in the context of dfBSD is that it's
used for their 1:1 threading model, which is different than the M:N
stuff that's being done recently. I think it's essential to think about
	I recall Rusty Russell added the Asynchronous interface to the
	Futex interface, yes, I am referring to the FUTEX_AWAIT and
	friends.  It was needed for the NGPT work, since it is M:N
	effectively.
	Regards,

--
Hiten Pandya
hmp at xxxxxxxxxxxxx